Harmonic Security raises $50M: know what your team pastes into AI

Harmonic Security raised $50 million to watch what employees do with AI, and that tells you where the risk is. The workforce AI security startup closed a Series B led by Crosspoint Capital Partners on October 14. Its product inspects employee use of AI tools in browsers, desktop apps and agentic workflows. The enterprise market pays for that. A ten-person shop needs the same answer at a smaller price: what data goes into which AI tool, and from whose account.

What actually happened

Per SecurityWeek:

• The round: $50 million Series B, led by Crosspoint, with existing investors in. Total funding is now $76 million.
• The company: Founded in 2023. Alastair Paterson is CEO and Bryan Woolgar-O'Neil is CTO.
• The product: It classifies each AI task, identifies how it is used and gives real-time control. It flags risks such as personal accounts and sensitive data exposure. Small language models analyze each interaction in context.
• Usage Intelligence: The same engine reports which AI use cases actually improve workflows.
• Use of funds: More engineering and go-to-market staff in the US and UK.

Other coverage says Harmonic claims more than 100 customer companies and an integration with Anthropic's Claude Compliance API. Those are company claims. Valuation and revenue were not disclosed.

Why it matters for your business

The risk Harmonic sells against is simple: an employee pastes a customer list, a contract or a payroll export into a free AI account that the company does not control. We wrote about the same problem when Akamai found 6% of AI chats carry company data. The second risk is the "personal accounts" flag. A personal account means your data sits under someone else's login, with someone else's retention settings.

Most small teams do not need a $50M platform. They need four things:

• One approved AI account per person, on a business plan. Business tiers give you admin control and data terms. Personal accounts give you neither.
• A one-page AI use policy. List what never goes into AI: card numbers, customer PII, passwords, unsigned contracts.
• SSO where you can. When someone leaves, their AI access leaves with them.
• A usage review each quarter. Ask what people actually use AI for, and keep the uses that save hours.

Then, if you grow past what a policy can hold, you buy the tool.

Sources: SecurityWeek.